Twitter has an antisemitism problem, and it’s not doing nearly enough to combat it, according to an investigation by the Anti-Defamation League.

“Twitter’s Failure to Enforce its Policy Against Antisemitism,” the title of a statement published by the ADL on July 14, cites 225 tweets posted during a nine-week period in early 2022 that expressed anti-Jewish sentiment and that repeated antisemitic tropes, including conspiracy theories about Jewish power and greed, Holocaust denial and accusations of pedophilia.

After ADL reported the offending tweets to San Francisco–based Twitter, only 11 of them, or 5 percent of the total, were taken down by the platform, the agency said. That ratio is not good enough for ADL, which closely monitors antisemitic hate speech and acts of violence against Jews and others.

“Twitter must enact its most severe consequences,” the report read, “and remove destructive, hateful content when reported by experts from the communities most impacted by such content.”

Twitter did not respond to a request for comment.

The ADL’s Silicon Valley–based Center for Technology and Society undertook the investigation, which retrieved 1 percent of all content on Twitter during a 24-hour window twice a week from Feb. 18 to April 21. The CTS filtered that content through its Online Hate Index, an algorithm designed to sift quickly through millions of posts and search for inarguably antisemitic hate speech. Human experts further narrowed the search down to 225 tweets that met the criteria.

One of those tweets said “You are owned by jews and I can’t imagine worse masters.” Another said, “It’s not a demon who wants a pound of flesh, it’s a jew.” Yet another said, in part, “If anyone is going to hell, it’s the last few generations of Judeo Americans …”

Wrote ADL: “As of May 24, 2022, 166 tweets of the 225 we initially reported to Twitter remain active on the platform. The remaining tweets have a potential reach of 254,455 users. While we know that Twitter removed 11 of the original 225 due to our directly reporting them, the reasons behind the removal of the other 48 are less clear.”

Twitter told the ADL that some tweets were “de-amplified” (meaning they could not be shared or engaged with) and that others “did not meet a particular threshold of hateful content on the platform,” according to the ADL. The statement also noted Twitter claimed that “tweets with one hateful comment were not subject to content moderation action … but if a tweet had repeated hateful comments, it would be subject to content moderation.”

Seth Brysk, the director of ADL’s S.F.-based Central Pacific Region, said the ADL was “very careful to include not only a cross-section of tweets, but we made sure these were the most egregious expressions of antisemitism. They were easy to spot.”

RELATED: ADL develops algorithm to track antisemitism on social media

Brysk said the tweets contained “classic expressions of Jewish power and greed, and hate directed at other communities, and yet to this day, over three-quarters of them are still on the platform, in violation of their own policy.”

Twitter policy, as stated in its “Abusive Behavior” section at help.twitter.com, expressly forbids threats of violence, Holocaust denial and the use of “insults, profanity, or slurs with the purpose of harassing or intimidating others.” Though the policy also notes that “while some individuals may find certain terms to be offensive, we will not take action against every instance where insulting terms are used.”

Consequences of violations include downranking and de-amplifying tweets, requiring tweet removal and banning violators from the platform.

The ADL decried Twitter’s response, noting that “the decisions made by Twitter significantly minimize the impact of antisemitism and that hate more broadly has on individuals from targeted communities. It is neither enough to de-amplify hate nor wait for hate to rise to a certain threshold before taking action. If a swastika was painted on a public building, we would not tell a community to put up a sign in front of it telling passersby to avert their eyes or wait to take action until there were a few additional swastikas painted on the same building.”

Brysk conceded that “hate speech is not a precise term” but added that “there are policies among these companies that largely address some of these issues, but sometimes they don’t get it right.

Only recently did Facebook start considering Holocaust denial as an “expression of hate,” he said. “Social media companies need to be willing to center the experiences of communities targeted by hate. There may be expressions of hatred directed at the Jewish community that might not be fully understood by some.”

Brysk echoed ADL’s recommendations that social media companies boost enforcement of their anti-hate-speech terms of use, strive for greater transparency around these issues and place greater trust in input from targeted communities.

He warned of the dangers of allowing hate speech to proliferate on social media.

When “denigrating and demonizing [groups and] referring to them as vermin, a cancer, and cockroaches” becomes normalized, that “starts to prepare [citizens] for the notion of eliminating these groups,” Brysk said.